AI Safety Fundamentals: Alignment
Un pódcast de BlueDot Impact
Categorías:
83 Episodo
-
Is Power-Seeking AI an Existential Risk?
Publicado: 13/5/2023 -
Where I Agree and Disagree with Eliezer
Publicado: 13/5/2023 -
Supervising Strong Learners by Amplifying Weak Experts
Publicado: 13/5/2023 -
Measuring Progress on Scalable Oversight for Large Language Models
Publicado: 13/5/2023 -
Least-To-Most Prompting Enables Complex Reasoning in Large Language Models
Publicado: 13/5/2023 -
Summarizing Books With Human Feedback
Publicado: 13/5/2023 -
Takeaways From Our Robust Injury Classifier Project [Redwood Research]
Publicado: 13/5/2023 -
AI Safety via Debatered Teaming Language Models With Language Models
Publicado: 13/5/2023 -
High-Stakes Alignment via Adversarial Training [Redwood Research Report]
Publicado: 13/5/2023 -
AI Safety via Debate
Publicado: 13/5/2023 -
Robust Feature-Level Adversaries Are Interpretability Tools
Publicado: 13/5/2023 -
Introduction to Logical Decision Theory for Computer Scientists
Publicado: 13/5/2023 -
Debate Update: Obfuscated Arguments Problem
Publicado: 13/5/2023 -
Discovering Latent Knowledge in Language Models Without Supervision
Publicado: 13/5/2023 -
Feature Visualization
Publicado: 13/5/2023 -
Toy Models of Superposition
Publicado: 13/5/2023 -
Understanding Intermediate Layers Using Linear Classifier Probes
Publicado: 13/5/2023 -
Acquisition of Chess Knowledge in Alphazero
Publicado: 13/5/2023 -
Careers in Alignment
Publicado: 13/5/2023 -
Embedded Agents
Publicado: 13/5/2023
Listen to resources from the AI Safety Fundamentals: Alignment course!https://aisafetyfundamentals.com/alignment